AI news · October 8, 2026
Google opens SynthID Detector for public watermark checks
SynthID Detector can check supported AI media from Google and partners, but a negative result is not proof that media is human-made. Watermark coverage remains the key limit.
- Google-stated watermarked images and videos
- 180B
- Google-stated watermarked audio
- 240,000 years
- daily verification requests in listed products
- 1M+
Google opened SynthID Detector globally in English on October 7. The service checks images, video, and audio for the SynthID watermark used by Google and participating partners, including OpenAI, NVIDIA, and Kakao, with Apple listed as a future partner. Google says more than 180 billion images and videos and the equivalent of 240,000 years of audio have been watermarked, and that its products already receive more than one million verification requests per day.
The product is useful as a quick check when the source tool is known, but it cannot establish that unmarked media is real. A detector can miss content that was made by a system without SynthID or changed after generation. Teams using it for moderation or publishing should record the source, keep the original file, and combine the result with human review and other evidence rather than presenting a score as certainty.
What you can do with it
Try SynthID Detector with a small set of known Google-generated and ordinary media. Add resize, recompression, and editing tests, then record false negatives and false positives. Use the result as one signal in a review flow, not as an automatic release or removal rule.
Our take
A public checker makes watermark research easier to test, and partner coverage gives the service a chance to become a shared utility. The important limitation is simple: it only detects marks that exist and survive the file changes. That makes source records and review steps necessary.
Links Try today Announcement
Source: Google DeepMind ↗ — Made With Models writes the brief; the reporting is theirs.